FIX Ensure everything gets flushed when flushing from sake

[GuySartorelli]

Ensures Flushable classes get flushed when flush booting the kernel, regardless of where that's happening from.

Issue

• CMS6: Flushable classes aren't flushed when flushing from sake #11435

[GuySartorelli]

That method is part of the Kernel interface, so no need for the old method_exists check.

[GuySartorelli]

This is the exact test (with some code quality improvements) that used to live in FlushMiddlewareTest

[GuySartorelli]

CI failures are identical to what's happening on the 6 branch - see https://github.com/silverstripe/silverstripe-framework/actions/runs/11448877901/job/31853334275

[emteknetnz]

Code is fine, though FlushMiddleware just seems pointless now, just delete it and put the following in HTTPCacheControlMiddleware::process()

$kernel = Injector::inst()->get(Kernel::class);
if ($kernel->isFlushed()) {
  $this->disableCache(true);
}

And deprecate in CMS 5.4.0, changelog entry, remove from yml, etc

Note there's also some docs that need to updated regardless, e.g. 01_Flushable.md is incorrect regardless if we remove FlushMiddleware or not, so do a search in the codebase for "FlushMiddleware"

Also I don't want to merge anything else into CMS 6 until we've fixed up all the broken builds, once that's done could you please re-run tests to confirm that things are green

[GuySartorelli]

once that's done could you please re-run tests to confirm that things are green

I assume you mean in this PR specifically lol

[GuySartorelli]

just delete it ... and deprecate in CMS 5.4.0, changelog entry, remove from yml, etc

I was avoiding unnecessary breaking changes, because this was meant to be just a quick bugfix. I'll put this back in ready, I've got way too many things running concurrently right now to do a full breaking change set of PRs for this.

[GuySartorelli]

Updated, CMS 5 PRs incoming